Professional Documents
Culture Documents
Abstract— Identification of masked objects especially in detection segmentation for IR images. Section 3 converses the
of landmines is always a difficult problem due to environmental comparison of segmentation methods and the subjective
inference. Here, segmentation phase is highly concentrated by assessment of the approach Section 4 explores the performance
performing an initial spatial segmentation to achieve a minimal evaluation of the methods. The paper ends with observations
number of segmented regions while preserving the homogeneity
criteria of each region. This paper aims in evaluating similarities
on future work and some conclusions.
based segmentation methods to compose the partition of objects in
Infra-Red images. The output is a set of non-overlapping II. PRELIMINARIES OF SEGMENTATION
homogenous regions that compose the pixels of the image. These
extracted regions are used as the initial data structure in feature There are many unsupervised and supervised segmentation
extraction process. Experimental results conclude that h-maxima algorithms [6]. They only use low-level features, e.g. intensity
transformation provides better results for landmine detection by and texture, to generate homogeneous patches from an input
taking the advantage of the threshold. The relative performance of image. Four categories for segmentation are: histogram shape
different conventional methods and proposed method are based methods, where, for example, the peaks, valleys and
evaluated and compared using the Global Consistency Error and curvatures of the smoothed histogram[4]. Clustering based
Structural Content. It proves that h-maxima gives significant methods, where the gray level samples are clustered in two
results that definitely facilitate the landmine classification system parts as background and foreground (object), or alternately are
more effectively.
modeled as a mixture of two Gaussians. Entropy based
Keywords- Segmentation, Global Consistency error, h-maxima, methods results in algorithms that use the entropy of the
threshold, Landmine detection foreground and background regions, the cross entropy between
the original and binarized image, etc. Object attribute-based
I. INTRODUCTION methods search a measure of similarity between the gray level
and the binarized images, such as fuzzy shape similarity, edge
The signature of buried land mine in IR images varies coincide, etc. This paper quantifies the segmentation methods
significantly depending on external parameters such as based on the similarities of the pixels including the intensity
weather, soil moisture, solar radiation, burial depth, and and the object structure.
time[7]. By literature [Oscar González Merino] [11] the
working of many image-based landmine detection algorithms,
it is concluded that the fundamental challenges arise from the III. SIMILARITIES BASED SEGMENTATION TECHNIQUES
fact that the mean spectral signatures of disturbed soil areas Segmentation is a pre-process which partitioned image into
that indicate mine presence are nearly always very similar to unique multiple regions, where region is set of pixels.
the signatures of mixed background pixels that naturally occur Mathematically segmentation can be defined as follows:
in heterogeneous scenes composed of various types of soil and
vegetation [1][12]. Mine detection using infrared techniques is If I is set of all image pixels, then by applying segmentation
primarily based on exploiting temperature differences between we get different unique regions like {S1, S2, S3,…,Sn } which
pixels on the mines and background pixels [2]. Thus, there is when combined formed „I‟ . Basic formulation is as follows:
always a need for robust algorithm that has the capability to n
analyze the pattern of distribution of the pixels to separate
(a) Si I where Si Sj
pixels of the mines from background pixels. Here the i 1, n
segmentation methods are evaluated with performance metrics
(b) Si is a connected region, i=1, 2….n.
[3]. Experimental results show that h-maxima transformation is
(c) P(S i) = TRUE for i=1, 2... n.
more adoptable for IR target images. The paper is organized as
follows: Section 2 deals with the need for pixel based (d) P Si S j FALSE fori j.
88 | P a g e
http://ijacsa.thesai.org
(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 1, No. 5, November 2010
1
mi( t 1)
Si( t )
X j -
x j S ( t ) i
-(2)
C. Threshold intensity distribution algorithm
It can detect object boundaries with low gradient or
reduce noise effect in gradient. However, an accurate and
stable estimation of intensity distribution is difficult to get
Figure 1: Schematic representation of proposed algorithm for identification of
from a finite set of 3D image data. To reduce the ”shrink” or
landmine ”expand” effect on segmentation results, gradient information
is used to calibrate the estimation of intensity distribution in
A. H-maxima algorithm the following. Overlap of image gradient is computed with
Every IR image taken has been implemented with the boundaries determined by intensity distribution through
morphological reconstruction, extended maxima transformation introducing a probability offset to intensity distribution. The
using thresholding[10] . The extended maxima transformation maximum overlap indicates the optimal boundaries of the
is the regional maxima computation of the corresponding h- interested objects. To restate the problem without losing
maxima transformation. As a result, it produces a binary image. generality, here it use a mixed Gaussian distribution model.
A connected-component labeling operation is performed, in n
P(u ) k P(u k ; k , k ),
order to evaluate the characteristics and the location of every
object. As a second object reduction step, objects not located
within a region of another object, are also discarded, since mine k 1
objects are not typically clustered. The region of interest (ie)
the target is got by connected component segmentation in
which the relevant pixels of the object will be grouped and
89 | P a g e
http://ijacsa.thesai.org
(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 1, No. 5, November 2010
k 1
k 1, and Rules that are worked out to guide the process of true
corner point localization.
Gaussian distribution of the intensity. Intensity distribution (i) Select those boundary points which bear significantly large
inside the region Ω is in Eq below cornerity index by eliminating the boundary points which lie
on straight line segments bearing negligibly small cornerity
P u ; ,
index value.
Pin (u ) k k k k (ii) Since all points on a smooth curve segment are in general
k
k associated with almost same cornerity index, and actual corner
And the intensity distribution of the outside region points bear cornerity index larger than that of their neighbors,
it is suggested to select the set of connected points such that
Ω, Pout , can be obtained in a similar way. Normally for pixel the variations in their cornerity indices are considerably large.
x on region boundaries with u = I(x), This rule helps in selecting only the set of points with in the
vicinity of actual corner points by eliminating the points on
Pout I ( x) Pin I ( x) 0 smooth curve segments.
(iii) Select the points which bear local maximum cornerity
Threshold Intensity Distribution is a significant aspect in index as true corner points. It could be noticed that these rules
assigning an individual region around an image[4]. Uneven do not require any priori knowledge in locating true corner
distribution may lead to assignment of two or more regions to points.
an individual pixel. This method reduces processing time by Thus, the expected point corresponding to a corner point
performing gray-level based segmentation that extracts regions will have a larger shift when compared to other points on the
of uniform intensity. Subsequently, it is also possible to boundary curve. Therefore, the cornerity index of pi is defined
estimate motion for the regions. It also reduces the to be the Euclidean distance d between the points pi and its
computational load, and the region-based estimator gives expected point pie and is given by
robustness to noise and changes of illumination. The
segmentation of the reference image is designed to group pixels
d xi xie yi yie
2 2
of similar gray-levels. In intensity distribution, B (i,j) is a
binary image (pixel are either 0 or 1) created by thresholding
F(i,j) The cornerity index indicates the prominence of a corner
point. The larger the value of the cornerity index of a boundary
B(i,j) = 1 if F (i,j)<t point, the stronger is the evidence that the boundary point is a
B (i,j) =0 if F(I,j)>=t corner.
It is assumed that the 1‟s are the object pixels and the 0‟s
are the background pixels.
The Histogram (h) - gray level frequency distribution of the
gray level image F.
90 | P a g e
http://ijacsa.thesai.org
(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 1, No. 5, November 2010
1.2
Range of values
1 boundary
0.8
intensity
0.6
kmeans
0.4
0.2 h-transform
0
e1
e2
e3
e4
e5
e6
e7
e8
e9
0
e1
ag
ag
ag
ag
ag
ag
ag
ag
ag
ag
Im
Im
Im
Im
Im
Im
Im
Im
Im
Im
Images taken
Structural Content
35
Range of values
30
boundary
25
20 intensity
15 kmeans
10
h transform
5
0
e1
e2
e3
e4
e5
e6
e7
e8
e9
0
e1
ag
ag
ag
ag
ag
ag
ag
ag
ag
ag
Im
Im
Im
Im
Im
Im
Im
Im
Im
Im
Images Taken
Figure 2: Visual assessment of the segmentation methods Figure 4: Comparison based on structural content
91 | P a g e
http://ijacsa.thesai.org
(IJACSA) International Journal of Advanced Computer Science and Applications,
Vol. 1, No. 5, November 2010
From the above figures 3, 4 the objective evaluation also [7] F. Cremer and W. de Jong and K. Schutte.„Infrared polarization
states that h maxima transform is the desirable method for measurements of surface and buried anti-personnel landmines‟. In
Preprint Proc. SPIE Vol. 4394, Detection and Remediation Techniques
landmine detection. The Global consistency error rate should for Mines and Minelike Targets VI, Orlando FL, USA, Apr. 2001.
always be low and it is the same for the proposed method in [8] Yan Sun, Chengyi Sun and Wanzhen Wang, “Color Images
comparatively to the conventional methods. The structural Segmentation Using New Definition of Connected Components”,
content is high for h-maxima which state that the originality of Proceedings of ICSP2000.
the object and its information remains the same after the [9] K.S. Ravichandran,B. Ananthi, “Color Skin Segmentation Using K-
segmentation process. Means Cluster” International Journal of Computational and Applied
Mathematics ISSN 1819-4966 Volume 4 Number 2 (2009), pp. 153–157
[10] Punam K. Saha, “Optimum Image Thresholding via Class Uncertainty
V. CONCLUSION and Region Homogeneity”,IEEE transactions on pattern analysis and
The method described in this paper provides a relatively machine intelligence, vol. 23, no. 7, july 2001
simple, extremely fast, and robust method for displaying and [11] Oscar González Merino “Image analysis of Infrared Polarization
measurements of Landmines”, Academic year: 2000/01
performing automatic target identification phase. The result of
image segmentation is a set of segments that collectively cover [12] H. Brunzell, “Clutter Reduction and Object Detection in Surface
Penetrating Radar,” in Proc. Radar.97 Conf., 1997 pp. 688-691.
the entire image, or a set of contours extracted from the image.
Each of the pixels in a region is similar with respect to some
AUTHORS PROFILE
characteristic or computed property, such as color, intensity, or
texture. Adjacent regions are significantly different with
respect to the same characteristic(s). From this paper it is Dr. Padmavathi Ganapathi is the Professor and
concluded that h-maxima is very useful in IR image evaluation Head of Department of Computer Science,
Avinashilingam University for Women,
but that it should probably restrict studies to similar images and Coimbatore. She has 23 years of teaching
similar processing. Simulations are carried out which experience and one year Industrial experience.
demonstrates the proposed method is able to successfully Her areas of interest include Network security and
localize landmine objects from different sets of real IR images. Cryptography and real time communication. She
Nevertheless, this scheme has some limitations because it is not has more than 108 publications at national and
International level. She is a life member of many
automatic as different parameters have to be adjusted manually. professional organizations like CSI, ISTE, AACE,
Future work should incorporate the use of high-level image WSEAS, ISCA, and UWA.
analysis methods for the identification of the true mine objects
among the set of the detected mine cues. Dr. Subashini is the Associate professor in
Department of Computer Science, Avinashilingam
Deemed University for Women, Coimbatore. She
REFERENCES has 16 years of teaching experience. Her areas of
[1] Frank cremer't, wim de jong and klamer schutte ,” processing of interest include Object oriented technology, Data
polarimetric infrared images for landmine detection”,2nd international mining, Image processing, Pattern recognition.
workshop on advanced gpr. 14-16 may, 2003, the Netherlands She has 55 publications at national and
[2] E Ctrmn, W. de long, and K Schutte, “Fusion of polarimetric infrared International level.
features and GPR features far landmine detection,” in 2nd International
Workshop on Advanced Ground Penetrating Radar (IWAGPR), Delft,
The Netherlands, May 2003.
[3] Nathir A. Rawashdeh, Shaun T. Love ” Hierarchical Image Ms.M.Krishnaveni has 4 Years of Research
Segmentation by Structural Content” journal of software, vol. 3, no. 2, Experience Working as Research staff in DRDO
february 2008 and pursuing her Ph.D in Avinashilingam
University for Women, Coimbatore. Her research
[4] Yongsheng Hu, Qian Chen,”An Infrared Image Preprocessing Method interest are Image Processing, Pattern
Using Improved Adaptive Neighborhoods “,Proceedings of the 6th Recognition and Neural Networks.She has 24
World Congress on Intelligent Control and Automation, June 21 - 23, publications at national and international level.
2006, Dalian, China
[5] Zhou Wang, Alan Conard Bovik, Hamid Rahim Sheik and Erno P
Simoncelli, “Image Quality Assessment: From Error Visibility to
Structural Similarity”, IEEE Trans. Image Processing, and Vol. 13,
(2004).
[6] R.M. Haralick, and L.G. Shapiro, “Survey: Image Segmentation
Techniques,” Computer Vision, Graphics, and Image Processing, Vol.
29, pp. 100-132, 1985.
92 | P a g e
http://ijacsa.thesai.org